Hello, friend.

Lucy Li


I'm a senior and master's student at Stanford, studying artificial intelligence as well as human minds, behavior, and language. I'm mainly centered around natural language processing (NLP), but also exploring related fields.

My current research projects use linguistic and social network analysis to better understand people's interactions with each other. Last summer I participated in CURIS with the SNAP Group, and I'm currently collaborating with University of Michigan professor David Jurgens.

I'm very involved with the Symbolic Systems Program as a board member of the student society . I enjoy reading, writing, boba, indoor gardening, and photography. I'm also passionate about education, especially when it comes to increasing diversity in science, technology, engineering, and math (STEM)1.

Wishlist: cool data visualizations, book recommendations, and cat papers.

Papers

I publish under the name L. Lucy. (The title for this section is misleading because I have only one paper right now... I'm working on fixing this, I promise.)

Are distributional representations ready for the real world? Evaluating word vectors for grounded perceptual meaning.
Li Lucy, Jon Gauthier.
Accepted to RoboNLP @ ACL 2017.
Widely-used word embeddings such as GloVe and word2vec may be deficient for predicting some types of meaning. Link.

Projects

Here are some of my course projects:

Paying Attention to SQuAD: Exploring Bidirectional Attention Flow
Heather Blundell, Lucy Li
This reimplementation of BiDAF was pretty close to the original paper: 76.037 F1 and 66.663 EM. Link.


Historical and Modern Image-to-Image Translation with Generative Adversarial Networks
Vinson Luo, Michael Straka, Lucy Li
Since everyone is doing GANs stuff these days... we used CycleGANs to colorize historical images. Link.


The Effect of Speech Disfluencies on Turn-Taking
Lucy Li, Kartik Sawhney, Divya Saini
We predicted end-of-turns in spoken conversations using filler words and discourse markers. Link.


A Resource Rational Analysis of Memory Use in Bayesian Updating
Lucy Li, George Pakapol Supaniratisai, Stephanie Zhang
If noise-less memories are costly, how should one update one’s beliefs when given new observations? We built a hierarchical Bayesian model of memory. Link.


DominAI: An AI Player for Dominoes
Guillermo Angeris, Lucy Li
We created an agent using game playing and sampling algorithms to beat opponents at four-player dominoes. Link.